Skip to main content

AI Assistant Quick Start

The AWSops AI Assistant is powered by Amazon Bedrock AgentCore, allowing you to ask questions and request analysis about your AWS infrastructure using natural language.

Getting Started​

1. Access the AI Assistant Page​

Click AI Assistant in the sidebar.

2. Enter Your Question​

Type your question in the input field at the bottom of the screen and press Enter or click the send button.

3. View the Response​

The AI analyzes your question, routes it to the appropriate Gateway, and generates a response using available tools.

Example Questions​

Resource Status Queries​

Show me the EC2 instance status
How many S3 buckets do we have?
List the Lambda functions

Network Analysis​

Analyze the VPC network configuration
Check if any security groups have 0.0.0.0/0 inbound rules

Security Checks​

Check for security issues
Are there any IAM users without MFA enabled?

Cost Analysis​

Show me this month's cost breakdown
Compare costs by service

Container Status​

What's the EKS cluster status?
Check the ECS service status

11-Route Classification​

The AI Assistant analyzes questions and automatically routes them to the most appropriate of 11 specialized routes.

PriorityRoutePurpose
1codePython code execution, calculations, visualization
2networkVPC, TGW, VPN, Flow Logs analysis
3containerEKS, ECS, Istio troubleshooting
4iacCDK, CloudFormation, Terraform
5dataDynamoDB, RDS, ElastiCache, MSK
6securityIAM, policy simulation, security summary
7monitoringCloudWatch, CloudTrail
8costCost analysis, forecasting, budgets
9datasourceExternal observability — Prometheus/Loki/Tempo/ClickHouse/Jaeger/Dynatrace/Datadog (natural language → query)
10aws-dataResource lists/status (Steampipe SQL)
11generalGeneral AWS questions, documentation search
Checking the Route

You can see which Gateway was used in the route information displayed at the bottom of each response. Example: Network Gateway (17 tools), Bedrock + Steampipe SQL

Understanding Responses​

Response Structure​

┌────────────────────────────────────────────────────────┐
│  [AI Icon]                                             │
│                                                        │
│  Response content (Markdown format)                    │
│  - Supports tables, lists, code blocks                 │
│                                                        │
├────────────────────────────────────────────────────────┤
│  Network Gateway (17 tools)  │  Claude sonnet-4.6  │ 3.2s │
├────────────────────────────────────────────────────────┤
│  Tools: list_vpcs, get_vpc_network_details, ...        │
│  Queried: aws_vpc, aws_vpc_subnet                      │
└────────────────────────────────────────────────────────┘

Displayed Information​

  • Route Path: Which Gateway processed the request
  • Model: The Claude model used (Sonnet/Opus)
  • Response Time: Time taken to process
  • Tools Used: List of MCP tools called
  • Queried Resources: Tables queried from Steampipe

Real-time Streaming​

Responses are streamed in real-time and displayed progressively on screen. The optimal streaming mode is automatically selected based on the response path:

  • Single Gateway response: Displayed naturally with a typing effect
  • Multi-route synthesis: Real-time streaming of synthesis results via Bedrock Converse API
  • Data queries (aws-data): Bedrock native token streaming

Model Selection​

Select a model from the dropdown in the upper right corner:

  • Claude Sonnet 4.6: Fast responses, suitable for general questions (default)
  • Claude Opus 4.8: Complex analysis, suitable for deep reasoning
  • Claude Haiku 4.5: Fast and low-cost, suitable for simple questions / high volume

After each response, related follow-up questions appear as buttons. Click one to automatically input that question.

Example:

[Show IAM user list with Access Key status]
[Check if any users don't have MFA enabled]
[Check for security groups with 0.0.0.0/0 inbound]

Conversation History​

Session History​

The current session's conversation is maintained on screen. You can ask follow-up questions that reference previous conversations.

Saved History​

Expand the Conversation History panel at the bottom of the screen to view and re-ask questions from previous sessions.

Next Steps​